Empirical comparison of niching methods on hybrid composition functions

In this paper, we compare the performance of three popular niching genetic algorithms namely deterministic crowding, restricted tournament selection, and clearing by a set of hybrid composition test functions originally proposed for the special session on real parameter optimization at CEC 2005. The number of function evaluations is used as the main control parameter for an unbiased comparison instead of using the generation count as done frequently in the previous comparative studies. Results are given in tables and graphs to show the searching ability, accuracy, and computation time requirement of each method.